WFGY/TensionUniverse/Experiments/Q125_MVP/README.md

TU Q125 MVP: toy multi agent AI dynamics

Status: work in progress. This page records early MVP designs and will be extended with concrete results later.

This page sketches simple multi agent experiments for TU Q125.
The aim is to make interaction tension visible in controlled toy setups.

Navigation

• ← Back to Experiments index
• ← Back to Event Horizon (WFGY 3.0)

---

0. What this page is about

TU Q125 looks at multi agent AI dynamics.

We work with:

• toy environments,
• several AI or scripted agents,
• interaction protocols.

The MVP experiments define observables tracking tension between:

• individual objectives,
• system level outcomes,
• and specified norms or safety rules.

---

1. Experiment A: shared resource with agent policies

1.1 Research question

In a simple shared resource environment, can we define a scalar observable T_multi that

• is small when agent policies coexist without collapse,
• grows when local optimization leads to depletion or conflict.

1.2 Setup

The notebook will:

• Define an environment with a renewable resource.

• Instantiate several agents with simple policies, such as:

  • greedy harvesters,
  • conservative harvesters,
  • rule following agents.

• Run repeated interaction episodes where:

  • agents choose actions,
  • resource regenerates or depletes,
  • payoffs are assigned.

Record:

• resource level over time,
• agent payoffs,
• violations of any shared rules.

Define T_multi from:

• long run resource depletion,
• inequality or instability in payoffs,
• number of rule violations.

1.3 Expected pattern

We expect:

• low T_multi when agent mix and policies maintain the resource,
• higher T_multi when interactions drive collapse or large instability.

1.4 How to reproduce

After Q125_A.ipynb exists:

1. Open the notebook.
2. Inspect the environment and policy definitions.
3. Run simulations with different agent mixes.
4. Compare T_multi across setups.

---

2. Experiment B: communication and miscoordination

2.1 Research question

What happens when agents can communicate, and can we define T_comm to capture miscoordination and deception tension.

2.2 Setup

The notebook will extend Experiment A by adding:

• a simple communication channel where agents send short messages,
• a protocol where agents can coordinate or mislead.

For each episode record:

• messages sent,
• actions taken,
• whether communication improved or harmed outcomes.

Define T_comm from:

• cases where communication increases T_multi,
• mismatch between stated intentions and observed actions.

2.3 Expected pattern

We expect:

• low T_comm when communication supports stable cooperation,
• higher T_comm when communication is used for exploitation or creates confusion.

2.4 How to reproduce

Once Q125_B.ipynb exists:

• open the notebook and inspect the communication model,
• run simulations with and without communication,
• compare T_comm and T_multi.

---

3. How this MVP fits into Tension Universe

TU Q125 treats multi agent AI dynamics as a tension between:

• local objectives,
• shared resources and norms,
• communication and coordination.

This MVP gives:

• a shared resource experiment with T_multi,
• a communication experiment with T_comm.

Both are intended as transparent starting points, not full simulations.

For overall context:

• Experiments index
• Event Horizon (WFGY 3.0)

---

Charters and formal context

This page follows:

• TU Effective Layer Charter
• TU Encoding and Fairness Charter
• TU Tension Scale Charter